It should be noted that a partial discharge is an electric discharge limited to a portion of the insulation of an electrical system and does not therefore cause immediate failure of the system but its gradual degradation. By their very nature, therefore, partial discharges are substantially limited in extent to a defect in the insulating system.
In light of this, diagnostic methods based on the detection and interpretation of partial discharges are among the most promising and widely studied in the context of scientific research since processing the signals relating to partial discharges makes it possible to investigate the nature of defects in the insulating system where the discharges occur and locate the position of the defects in the insulating system.
More specifically, this invention relates to the diagnostics of electrical apparatuses with elongate geometry constituting a line for the transmission of the discharge pulses, that is to say, forming a path along which the pulses propagate; for example, apparatuses of this kind include electrical cables for medium and high voltages, autotransformers or GILs (gas insulated lines).
In this type of apparatus, it is important to identify any defects that might lead to cable failure, such as, for example, defects in the joints or terminations or even in the cable insulation. These defects are usually the site of partial discharges; generally speaking, therefore, the object is to identify the defects by detecting the corresponding partial discharges using suitable sensors coupled to the cable being checked.
Several methods are known for locating a discharge site along a cable.
One of these, known as reflectometric method, involves acquiring the signals at a detecting station located at one end of the cable and measuring the time interval that elapses between one signal and the signal reflected back by that signal from the opposite end of the cable. The distance of the source of that signal from the detection point is then calculated on the basis of the speed at which the signals propagate in that cable.
This method is not very reliable for long cables since the signals travelling the cable are attenuated to such an extent that there is the risk of not being able to detect them at all at the detecting station.
Another method involves coupling to the cable a plurality of sensors (at least 2, and usually 3) at different positions along the cable (thereby forming a plurality of detecting stations). When the signal at one of the sensors exceeds a predetermined level, the signals from the different sensors are acquired synchronously; the acquisition time window must be long enough to allow measurement of the same pulse in transit through the different sensors. Comparing the times at which the same signal reaches each of the sensors makes it possible to locate the source of the signals along the cable.
This method is more accurate than the reflectometric method but poses some problems.
A first problem is that of identifying a pair of homologous pulses, that is to say, pulses assignable to the same partial discharge and propagating in opposite directions.
It should be noted that the presence of a pair of homologous pulses detected at two different sensors spaced from each other along the apparatus means that the site where the discharge that generated these pulses took place is located between the two sensors.
In light of this, the prior art solutions (for example, U.S. Pat. Nos. 5,070,537 or 6,366,095) teach synchronizing the signals detected at the different sensors through an absolute time reference, for example a GPS.
These prior art devices comprise:                a first and a second sensor connected to the apparatus in a first and a second detecting station and spaced out along the apparatus, for detecting electric signals;        a processing unit connected to the sensors for receiving the signals and having a module for selecting at least one pair of signals, detected in the first and second sensors and representative of a pair of homologous pulses (that is to say, pulses relating to the same partial discharge and propagating in opposite directions along the apparatus), and a module for calculating the distance between the discharge site and the detecting stations based on the selected pair of homologous pulses.        
In the prior art solutions, the calculation module processes the following information: an estimate of the speed of pulse propagation in the cable, the distance between the detecting stations and the time that elapses between the instants the pulses of the pair considered are detected.
From the time that elapses between the instants the pulses of the pair considered are detected is derived the quantity by which the discharge site is displaced along the cable in the direction of one of the detecting stations relative to the other. In light of this, if the signals of the pair are detected simultaneously, it means the discharge site is equidistant from the detecting stations.
Based on this information and on the distance between the detecting stations, the calculation module finds the distance of the discharge site from the detecting stations.
These systems have two drawbacks, however.
A first drawback is due to the difficulty of reliably identifying the partial discharges to be examined. That is to say, the fact that two signals are detected by respective sensors in a sufficiently short interval of time (compatibly with the estimated speed of pulse propagation in the apparatus and the distance between the sensors) does not guarantee that those signals are from partial discharges, and in particular from the same partial discharge.
In fact, numerous signals may be coupled to the sensors other than the target signals corresponding to the partial discharges to be identified; these unwanted signals may consist of background noise or disturbances of another nature, or even other partial discharges (occurring at another discharge site or outside the cable).
All these signals substantially accompany the target signals generated by the partial discharges to be detected and often prevent the latter from being identified (especially if their amplitude is greater than that of the target signals to be detected).
To overcome this drawback, the prior art methods teach the use of suitable adjusted filtering sensors and systems in an attempt to detect the partial discharge signals without detecting the noise signals.
These systems do not, however, make it possible to determine whether the signals detected are due to the same partial discharge activity.
Moreover, these methods have inherent shortcomings.
Indeed, it is impossible to provide sensors optimized for each and every circumstance and the filtering systems are often ineffective. Also, since the frequency of the target discharge signals to be detected is not known beforehand, there is the risk of filtering out those very signals.
Lastly, when the signals generated by target partial discharge activities to be detected are accompanied by other signals generated by partial discharges to be ignored (for example because they are outside the cable), there is the risk that these signals cannot be selectively processed with filtering systems currently in use (consisting, for example, of analog passband filters).
A second shortcoming of known locating systems relates to the accuracy with which the target is located from the selected pair of homologous pulses.
The above mentioned prior art solutions involve processing the time phase shift between the signals in the pair of homologous pulses. This time phase shift allows calculation of the distance of the sensors from the discharge site, based on an estimation of the pulse propagation speed when the distance between the sensors is known.
This time domain comparison of the signals implies identifying corresponding reference points in the two signals (since the two pulses are not ideal but have each its own far from negligible development in time).
Typically, the first peak (or the highest peak) of the pulse is used as the reference for processing the time phase shift of the pulses.
This may lead to serious processing errors since the pulses, during their propagation in the apparatus, are deformed in proportion to the space travelled.
Also, there is a limit to the precision with which a time phase shift can be sensed by the instruments (even on the absurd assumption that the reference points for the pulses to be compared match perfectly). For example, a time precision of 100 ns, assuming that the pulses propagate at the speed of light, translates to a precision of 30 m.
Thus, the difficulties and shortcomings of the prior art as regards the accuracy of locating the signals detected are added to those regarding the identification of the signals actually associated with the discharge activities and in particular with the same partial discharge.
In light of this, the prior art methods do not guarantee reliable results and are often ineffective.
Furthermore, it is known from patent document U.S. Pat. No. 5,719,503 a system for searching anomalies (such as surface corrosion) on conductive members such as pipes. Said anomalies are not associated to any electrical discharge and therefore do not generate any detectable pulse. Indeed, said conductive members are not subject to electrical voltage.
In this light, U.S. Pat. No. 5,719,503 teaches to generate two electrical pulses and to inject them in said conductive member at different sections, in such a way that these pulses intersect at predetermined locations along the conductor. These pulses travel along the conductive member, passing through the sections having the anomalies. Such pulses, modified as a result of inherent nonlinearities of said anomalies, are then detected; said modifications of the detected pulses are analyzed to localize the anomalies along the conductive member.
However, the teachings of U.S. Pat. No. 5,719,503 are not applicable to the localization along an electrical apparatus of partial discharges which take place in a defect (i.e. a discharge site) of the insulation of an electrical apparatus. In fact, the partial discharges themselves generate electrical pulses which originate in a single discharge site and propagate along the insulation of the electrical apparatus departing from the discharge site in opposite direction; indeed, partial discharges occur only when the electrical apparatus is subject to electrical voltage. Therefore, the injection of further pulses travelling in the conductor of the electrical apparatus would be useless with respect to the localization of the partial discharges.